Proc. Natl. Acad. Sci. USA Vol. 95, pp. 5690–5693, May 1998 Immunology

Natural killer-like nonspecific tumor cell lysis mediated by specific ligand-activated V␣14 NKT cells

TETSU KAWANO*, JUNQING CUI*, YASUHIKO KOEZUKA†,ISAO TOURA*, YOSHIKATSU KANEKO*, HIROSHI SATO*, EISUKE KONDO*, MICHISHIGE HARADA*, HARUHIKO KOSEKI*, TOSHINORI NAKAYAMA*, YUJIRO TANAKA*, AND MASARU TANIGUCHI*‡

*Core Research for Evolutional Science and Technology Project, Japan Science and Technology Corporation and Division of Molecular Immunology, Center for Biomedical Science, School of Medicine, Chiba University, 1–8-1 Inohana, Chuo, Chiba 260-8670, Japan; and †Pharmaceutical Research Laboratory, Kirin Brewery Co., Ltd., 3 Miyahara, Takasaki, Gunma 370-1202, Japan

Communicated by Kimishige Ishizaka, Kirin Brewery Co., LTD., Tokyo, Japan, March 4, 1998 (received for review January 6, 1998)

ABSTRACT We have recently identified ␣-galactosylcer- be ␣-galactosylceramide (␣-GalCer) (16). In addition, amide (␣-GalCer) as a specific ligand for an invariant V␣14͞ ␣-GalCer is presented by a monomorphic non-major- V␤8.2 receptor exclusively expressed on the majority of histocompatibility complex (non-MHC) class Ib mol- V␣14 NKT cells, a novel subset of lymphocytes. Here, we ecule, CD1d, expressed on dendritic cells (DC), and the report that ␣-GalCer selectively activates V␣14 NKT cells ligand͞CD1 complex selectively stimulates to proliferate resulting in prevention of tumor metastasis. The effector V␣14 NKT cells but not other lymphocytes only if costimu- mechanisms of the ligand-activated V␣14 NKT cells seem to latory signals generated by CD40͞CD40 ligand and B7͞ be mediated by natural killer (NK)-like nonspecific cytotox- CD28 interactions are provided (16). The results suggest that icity. Indeed, the cytotoxic index obtained by ␣-GalCer- activation mechanisms of V␣14 NKT cells are similar to activated V␣14 NKT cells was reduced by the addition of cold those of conventional T cells. However the effector mech- target tumor cells or by treatment with concanamycin A, anisms of V␣14 NKT cells after activation with ␣-GalCer which inhibits activation and secretion of perforin, but not by have not yet been determined. Here, we demonstrate that mAbs against molecules involved in the NKT cell recognition one of the effector functions of ␣-GalCer-activated V␣14 and conventional cytotoxicity, such as CD1d, V␤8, NK1.1, NKT cells is a target tumor cell lysis mediated by a TCR- Ly49C, Fas, or . These results suggest that the independent, NK-like nonspecific mechanism with perforin. ligand-activated V␣14 NKT cells kill tumor cells directly ͞ ␣ through a CD1d V 14 T cell receptor-independent, NK-like MATERIALS AND METHODS mechanism. Mice. V␣14 NKT-deficient (J␣281Ϫ͞Ϫ) mice were estab- ␣ A novel lymphoid lineage, V␣14 natural killer (NK) T cells, lished by specific deletion of the J 281 gene segment with distinct from other lymphoid cells including T cells, B cells, and homologous recombination and aggregation chimera tech- ␣ NK cells, is characterized by the early development at day 9.5 niques (14). In these mice, only V 14 NKT cells are not of gestation before thymus formation (1) and also by coex- generated, whereas other lymphoid populations, T cells, B ␣ pression of the NK receptor and a single, invariant T cell cells, and NK cells remain intact. The V 14 NKT-deficient ␣ Ϫ͞Ϫ ͞ receptor (TCR) encoded by V␣14 and J␣281 gene segments (J 281 ) mice were backcrossed three times with C57BL 6 ␣ Ϫ͞Ϫ͞ ␣ ␤ (2–4) in association with a highly skewed set of V␤s, mainly mice (14). V 14 NKT (RAG V 14tg V 8.2tg; tg, trans- ͞ V␤8.2 (5–13). Moreover, the generation of V␣14 NKT cells is genic) mice with a C57BL 6 background were established by Ϫ͞Ϫ͞ ␤ Ϫ͞Ϫ͞ ␣ exclusively dependent on the expression of the invariant V␣14 mating RAG V 8.2tg mice and RAG V 14tg mice as ␣ ␣␤ TCR as evidenced by the fact that V␣14 NKT cells do not described (16). In the V 14 NKT mice, only transgenic TCR ␣ ␤ develop in the invariant V␣14 TCR-deficient mice (14) and (V 14tg and V 8.2tg) are expressed and result in preferential ␣ that the forced expression of the invariant V␣14 TCR leads development of V 14 NKT cells with no detectable number of ͞ exclusively to V␣14 NKT cell generation and blocks conven- T cells or NK cells (16). Pathogen-free C57BL 6 mice were tional T cell development (15, 16). purchased from Japan SLC, Inc. (Shizuoka, Japan). All mice Although physiological functions of V␣14 NKT cells remain used in this study were maintained in specific pathogen-free to be elucidated, the extensive analysis has shown that V␣14 conditions. NKT cells are able to mediate allograft bone marrow rejection Evaluation of the Severity of Hepatic Metastasis. B16 b ϫ 6 (17), control autoimmune disease development (18, 19), and melanoma cells (H-2 ,3 10 per mouse) were injected into the spleen of young adult wild-type, V␣14 NKT-deficient, and produce large amounts of both (IL-4) and ␣ ␣ ␮ ͞ interferon ␥ (16, 20, 21). These findings suggest that V␣14- V 14 NKT mice. Then, -GalCer (100 g kg) or control NKT cells play complicated roles in regulating immune re- vehicle was injected intraperitoneally on days 1, 5, and 9. The sponses more than simply as IL-4 or interferon ␥ providers for treated mice were sacrificed on day 14, and their livers were Th2 or Th1 cell development, respectively, so far reported (22, photographed and subjected to quantitation of GM3 mela- 23). In addition, we have demonstrated recently that V␣14 noma (24–26). NKT cells are a primary target of IL-12 and exert a major Quantitation of GM3 Melanoma in Metastatic effector function in IL-12-mediated tumor rejection (14). Liver. For measurement of a melanoma antigen (GM3 gan- A ligand for the invariant V␣14͞V␤8.2 TCR exclusively glioside), each liver was homogenized as described (24, 25). expressed on V␣14 NKT cells has been identified recently to The homogenates of normal or metastatic livers (50 mg) or

␣ ␣ The publication costs of this article were defrayed in part by page charge Abbreviations: -GalCer, -galactosylceramide; TCR, T cell receptor; CMA, concanamycin A; tg, transgenic; NK, natural killer; MHC, payment. This article must therefore be hereby marked ‘‘advertisement’’ in major histocompatibility complex; E͞T, effector͞target; IL, interleu- accordance with 18 U.S.C. §1734 solely to indicate this fact. kin. © 1998 by The National Academy of Sciences 0027-8424͞98͞955690-4$2.00͞0 ‡To whom reprint requests should be addressed. e-mail: taniguti@ PNAS is available online at http:͞͞www.pnas.org. med.m.chiba-u.ac.jp.

5690 Downloaded by guest on September 30, 2021 Immunology: Kawano et al. Proc. Natl. Acad. Sci. USA 95 (1998) 5691

B16 melanoma cells were treated with 1 ml of extraction buffer containing 1% Nonidet P-40. Various concentrations of mel- anoma cell homogenates were used to make standard curves. GM3 melanoma antigens in the extracts were quantitated by RIA with 125I-labeled M2590 anti-melanoma GM3 mAb as described (24, 25). 51Cr Release Cytotoxicity Assay. Target cells were labeled with 100 ␮Ci sodium chromate (Amersham) per 5 ϫ 106 cells for 1 hr. ␣-GalCer-activated spleen cells from V␣14 NKT mice were used as effector cells and seeded into 96-well round- bottomed plates at the indicated effector͞target (E͞T) ratios on 51Cr-labeled B16 melanoma cells (1 ϫ 104). Radioactivity released from lysed target cells was counted on a ␥-counter after 4 hr of incubation at 37°C in 5% CO2. The percentage of specific 51Cr release was calculated by the following formula: % of specific lysis ϭ (sample cpm Ϫ spontaneous cpm) ϫ 100͞(maximum cpm Ϫ spontaneous cpm). Spontaneous cpm was calculated from the supernatant of the target cells alone, and the maximum release was obtained by adding 1 M HCl to target cells. The data are expressed as a mean value of FIG. 1. Inhibition of liver metastasis of B16 melanoma cells by in triplicate cultures with standard deviations. ␣ ␣ vivo treatment with -GalCer. (A) Melanoma antigens in metastasized For cold target inhibition experiments, a mixture of - livers measured by RIA using 125I-labeled M2590 mAb. B16 melanoma ␣ 51 GalCer-activated V 14 NKT cells and Cr-labeled B16 mel- cells (3 ϫ 106 per mouse) were inoculated in the spleen of wild-type anoma target cells (at an E͞T ratio of 25:1) was incubated with (J␣281ϩ͞ϩ), V␣14 NKT-deficient (J␣281Ϫ͞Ϫ), and V␣14 NKT unlabeled tumor cells (B16 or EL-4) at the indicated target͞ (RAGϪ͞Ϫ͞V␣14tg V␤8.2tg) mice, followed by intraperitoneal injec- inhibitor ratios. For antibody-blocking experiments, assays tion with either ␣-GalCer (100 ␮g͞kg) or vehicle on days 1, 5, and 9. were performed at an E͞T ratio of 50:1 in the presence of 50 The mice were sacrificed and examined on day 14. Each group ␮g͞ml of mAbs specific for H-2Kb (AF6–88.5), H-2Db (KH- consisted of three mice. The data are expressed as a mean cpm value 95), CD1d (1B1), V␤8 (MR5–2), NK1.1 (PK136), Ly49C of triplicate samples with standard errors. (B) Photographic views of metastasized livers from mice treated with ␣-GalCer or vehicle. (5E6), Fas (Jo2) (PharMingen), and Fas ligand (MFLIII was a kind gifted from H. Yagita, Juntendo University, Tokyo). ␣ ␣ administration of -GalCer suppressed the metastasis of B16 Treatment of V 14 NKT Cells with Concanamycin A melanoma cells as effectively as observed in wild-type mice. (CMA). ␣ V 14 NKT cells were preincubated with CMA (Wako Because the number of V␣14 NKT cells in the spleen of V␣14 Pure Chemical, Osaka) or control vehicle (dimethyl sulfoxide) NKT mice (7–9 ϫ 105) is similar to that in the wild-type mice at indicated concentrations as described (14, 27, 28). (6–9 ϫ 105), the efficacy of tumor-killing activity appears to be The assay was performed at an E͞T ratio of 50:1. Rausher B virus-induced T cell lymphomas (RMA-S) and their CD1d- equivalent. Fig. 1 shows representative photographic views of transfectants (RMA-S-CD1.1) were generated by L. Brossay, the metastasized livers. The metastatic melanoma nodules La Jolla Institute of Allergy and Immunology (La Jolla, CA). were visible as large black spots, which are not observed in the liver of either wild-type or V␣14 NKT mice if treated with ␣-GalCer. These results strongly suggest that the V␣14 NKT RESULTS AND DISCUSSION cells were selectively activated by ␣-GalCer and revealed an Antitumor Activity of V␣14 NKT Cells Activated by ␣- inhibitory effect on tumor metastasis in the liver. Similar in GalCer. The goal of this study was to clarify the effector vivo effects of ␣-GalCer-mediated protection of tumor metas- mechanisms of the cytotoxic function of ligand-activated V␣14 tasis were observed in other tumor systems, such as Lewis lung NKT cells. Recently, we have identified ␣-GalCer as a ligand carcinoma, erythroleukemia (FBL-3), Colon26, or RMA-S for the invariant V␣14 TCR exclusively expressed on V␣14 tested so far (data not shown). NKT cells (16). To investigate effector function of ␣-GalCer- As shown in Fig. 2, cytotoxic activity against B16 mela- activated V␣14 NKT cells, we used three different types of noma was detected in vitro when V␣14 NKT mice were mice: (i) wild-type mice bearing normal numbers of T cells, B injected with ␣-GalCer but not with vehicle, indicating that cells, NK cells, and V␣14 NKT cells, (ii)V␣14 NKT-deficient ␣-GalCer-activation is required for mediating effector func- mice that lack only V␣14 NKT cells while other lymphoid populations (T cells, B cells, and NK cells) remain intact (14), and (iii)V␣14 NKT mice in which only V␣14 NKT cells but not T cells, B cells, or NK cells are present (16). We injected ␣-GalCer intraperitoneally after intrasplenic inoculation of B16 melanoma cells. The severity of melanoma metastasis was evaluated by measuring amounts of GM3 melanoma antigen in the liver. As shown in Fig. 1A, a negligible level of GM3 melanoma antigen was detected when ␣-GalCer but not vehicle was administrated in wild-type mice, indicating that ␣-GalCer treatment prevented liver metastasis of B16 melanoma almost completely. In contrast, the prevention of metastasis was not observed in V␣14 NKT-deficient mice lacking only V␣14 NKT cells. Because T cells and NK cells other than V␣14 NKT cells are generated normally in V␣14 FIG. 2. Cytotoxicity of spleen cells from V␣14 NKT mice after in NKT-deficient mice (14), the antitumor effect by ␣-GalCer ␣ ␣ ␣ vivo treatment with -GalCer. V 14 NKT mice were injected intra- treatment appeared to be a consequence of the defect of V 14 peritoneally with either ␣-GalCer (100 ␮g͞kg) or vehicle, and 24 hr NKT cells. In V␣14 NKT mice bearing only V␣14 NKT cells later spleen cells were used as effector cells for cytotoxic assay on B16 without other lymphoid cells, as anticipated, the in vivo melanoma cells at indicated effector͞target ratios. Downloaded by guest on September 30, 2021 5692 Immunology: Kawano et al. Proc. Natl. Acad. Sci. USA 95 (1998)

tion of V␣14 NKT cells. These findings confirm the in vivo data on the rejection of melanoma after administration of ␣-GalCer. NK-Like Effector Function of ␣-GalCer-Activated V␣14 NKT Cells. Next, we performed a series of experiments aimed at elucidating the molecular mechanism involved in the effector cytotoxic function of the ␣-GalCer-activated V␣14 NKT cells. Fig. 3A demonstrates a representative result of a cold target inhibition assay on B16 melanoma cell killing by ␣-GalCer-activated V␣14 NKT cells. 51Cr release was reduced significantly by the addition of nonlabeled B16 melanoma cells in a dose-dependent manner, but not by ␣ EL-4, which is not susceptible to V 14 NKT cell-mediated FIG. 4. Molecular requirements in the effector phase of cytotoxic target cell lysis, indicating the requirement of direct cell-to- function of ␣-GalCer-activated V␣14 NKT cells. (A) Effects of various cell contact. In addition, the treatment of V␣14 NKT cells mAbs on V␣14 NKT cell-mediated cytotoxicity. Fifty micrograms per with CMA inhibited antitumor cytotoxic activity of the ml of the indicated mAbs were added to the cytotoxic assays. (B) ␣-GalCer-induced V␣14 NKT cells in a dose-dependent Cytotoxic activity of V␣14 NKT cells on ␣-GalCer-pulsed class I-deficient RMA-S cells and their CD1d-transfectants. RMA-S and fashion (Fig. 3B). Because CMA is a specific inhibitor of ␣ vacuolar-type Hϩ-ATPase and inhibits the perforin- RMA-S-CD1.1 cells were pulsed with -GalCer and used as target cells. The data are expressed as a mean value of triplicate cultures with mediated killing but not Fas-dependent cytotoxic pathway standard deviations. (27, 28), the effector function of V␣14 NKT cells appeared to be mediated by NK-like mechanisms. cytotoxic mechanisms of IL-12-activated V␣14 NKT cells We tried to identify the cell surface molecule essential for against tumor cells (14), suggesting that the ␣-GalCer- direct cell contact involved in the cytotoxic interactions by activated V␣14 NKT cells use a novel effector system using various mAbs with a specific blocking activity (Fig. mediating antitumor cytotoxicity. Although we could not 4A). Interestingly, however, no significant inhibition was block cytotoxic activity by the addition of antibodies against observed by the addition of mAbs specific for Fas ligand, Fas, interferon ␥ (data not shown) or Fas ligand (see Fig. 4A), it ␤ b b Ly49C, NK1.1, TCRV 8, CD1d, H-2D or H-2K . The data is still possible that some are involved in the shown in Fig. 4B in part confirm the results in Fig. 4A on the effector killing mechanisms. ͞ ␣ nonessential requirement of CD1d invariant V 14 TCR In the present study, we demonstrated a unique feature of interactions in the effector phase. The ␣-GalCer-activated ␣ molecular requirements in activation and effector function V 14 NKT cells killed both RMA-S and RMA-S-CD1.1 of V␣14 NKT cells. Similar to the activation of conventional cells, regardless of their CD1d expression or their antigen ␣ ␣ ͞ T cells, the specific ligand recognition by TCR is required for pulse with -GalCer, confirming that the invariant V 14 the activation of V␣14 NKT cells. The invariant V␣14 TCR V␤8.2 TCR is not used in the cytotoxic interaction between ␣ ␣ recognizes -GalCer presented by a CD1d molecule and V 14 NKT cells and tumor target cells. The system appeared mediates activation signals to V␣14 NKT cells together with not to involve either Ly49C͞NK1.1 molecules essential for costimulatory signals generated by B7͞CD28 and CD40͞ conventional NK cell-mediated killing or TCR͞MHC class I CD40 ligand interactions between dendritic cells and V␣14 molecules necessary for CTL-mediated killing mechanisms. NKT cells. Interestingly, however, the activated V␣14 NKT In particular, nonactivated V␣14 NKT but not NK cells cells kill tumor targets by NK-like mechanisms in a nonspe- killed FBL-3 tumor cells expressing MHC class I molecules cific TCR-independent fashion. Because ␣-GalCer so far has on the target cell surface (Fig. 5). It is well documented that NK cells killed only MHC class I negative target cells not been detected on tumor target cells, it is necessary to provide a ligand stimulation through dendritic cells to because of their negative signal through interaction of Ly49C ␣ molecule on NK cells with MHC class I molecule on target activate V 14 NKT cells. Once they are activated, they cells (29). Therefore, V␣14 NKT cells killed the target in the express nonspecific cytotoxic functions against a variety of tumors. In of the successful control of V␣14 NKT cell different mechanisms from those of NK-mediated cytolysis. ␣ These results indeed are in agreement with the effector activation in vivo, -GalCer is an ideal drug to manipulate tumor eradication.

FIG. 3. Effector mechanism of ␣-GalCer-activated V␣14 NKT cells. (A) ␣-GalCer-activated V␣14 NKT cells were incubated with FIG. 5. Cytotoxicity of NK and V␣14 NKT cells against FBL-3 51Cr-labeled B16 melanoma cells in the presence of various numbers erythroleukemia cells. (A) Fluorescence-activated cell sorter profiles of unlabeled tumor cells (B16 or EL-4) as cold targets. Cytotoxic assays of FBL-3. Cells were stained with anti-H-2Kb (AF6–88.5) or H-2Db were performed at an E͞T ratio of 25:1. (B) Effect of CMA treatment (KH95) antibodies. Isotype-matched antibody was used as a negative on ␣-GalCer-activated V␣14 NKT cells. ␣-GalCer-activated V␣14 control. (B) Cytotoxic activity. Freshly isolated spleen cells from V␣14 NKT cells were treated with indicated doses of CMA for 2 hr, and their NKT mice (F)orRAGϪ͞Ϫ (NK) mice (E) were assayed for their cytotoxic activity was assessed on 51Cr-labeled B16 melanoma at an cytotoxicity by the 4-hr 51Cr-release assay on FBL-3 tumor cells E͞T ratio of 50:1. expressing MHC class I molecules on the cell surface. Downloaded by guest on September 30, 2021 Immunology: Kawano et al. Proc. Natl. Acad. Sci. USA 95 (1998) 5693

We thank H. Tanabe for the preparation of this manuscript. Kondo, E., Kawano, T., Cui, J., Perkes, A., Koyasu, S. & Makino, Y. (1996) Proc. Natl. Acad. Sci. USA 93, 11025–11028. 1. Makino, Y., Kanno, R., Koseki, H. & Taniguchi, M. (1996) Proc. 16. Kawano, T., Cui, J., Koezuka, Y., Toura, I., Kaneko, Y., Motoki, Natl. Acad. Sci. USA 93, 6516–6520. K., Ueno, H., Nakagawa, R., Sato, H., Kondo, E., et al. (1997) 2. Imai, K., Kanno, M., Kimoto, H., Shigemoto, K., Yamamoto, S. Science 278, 1626–1629. & Taniguchi, M. (1986) Proc. Natl. Acad. Sci. USA 83, 8708– 17. Takeda, K. & Dennert, G. (1993) J. Exp. Med. 177, 155–164. 8712. 18. Mieza, M. A., Itoh, T., Cui, J. Q., Makino, Y., Kawano, T., 3. Lantz, O. & Bendelac, A. (1994) J. Exp. Med. 180, 1097–1106. Tsuchida, K., Koike, T., Shirai, T., Yagita, H., Matsuzawa, A., et 4. Makino, Y., Kanno, R., Ito, T., Higashino, K. & Taniguchi, M. al. (1996) J. Immunol. 156, 4035–4040. (1995) Int. Immunol. 7, 1157–1161. 19. Sumida, T., Sakamoto, A., Murata, H., Makino, Y., Takahashi, 5. Fowlkes, B. J., Kruisbeek, A. M., Ton-That, H., Weston, M. A., H., Yoshida, S., Nishioka, K., Iwamoto, I. & Taniguchi, M. (1995) Coligan, J. E., Schwartz, R. H. & Pardoll, D. M. (1987) Nature J. Exp. Med. 182, 1163–1168. (London) 329, 251–254. 20. Arase, H., Arase, N. & Saito, T. (1996) J. Exp. Med. 183, 6. Budd, R. C., Miescher, G. C., Howe, R. C., Lees, R. K., Bron, C. 2391–2396. & MacDonald, H. R. (1987) J. Exp. Med. 166, 577–582. 21. Yoshimoto, T. & Paul, W. E. (1994) J. Exp. Med. 179, 1285–1295. 7. Ballas, Z. K. & Rasmussen, W. (1990) J. Immunol. 145, 1039– 22. Yoshimoto, T., Bendelac, A., Watson, C., Hu-Li, J. & Paul, W. E. 1045. (1995) Science 270, 1845–1847. 8. Sykes, M. (1990) J. Immunol. 145, 3209–3215. 23. Yoshimoto, T., Bendelac, A., Hu-Li, J. & Paul, W. (1995) Proc. 9. Levitsky, H. I., Golumbek, P. T. & Pardoll, D. M. (1991) Natl. Acad. Sci. USA 92, 11931–11934. J. Immunol. 146, 1113–1117. 24. Wakabayashi, S., Okamoto, S. & Taniguchi, M. (1984) Gann 75, 10. Takahama, Y., Kosugi, A. & Singer, A. (1991) J. Immunol. 146, 427–432. 1134–1141. 25. Gunji, Y. & Taniguchi, M. (1986) Jpn. J. Cancer Res. 77, 595–601. 11. Bix, M., Coles, M. & Raulet, D. (1993) J. Exp. Med. 178, 901–908. 26. Nores, G. A., Dohi, T., Taniguchi, M. & Hakomori, S. (1987) 12. Bendelac, A., Killeen, N., Littman, D. R. & Schwartz, R. H. J. Immunol. 139, 3171–3176. (1994) Science 263, 1774–1778. 27. Kataoka, T., Shinohara, N., Takayama, H., Takaku, K., Kondo, 13. Masuda, K., Makino, Y., Cui, J., Ito, T., Tokuhisa, T., Takahama, S., Yonehara, S. & Nagai, K. (1996) J. Immunol. 156, 3678– Y., Koseki, H., Tsuchida, K., Koike, T., Moriya, H., et al. (1997) 3686. J. Immunol. 158, 2076–2082. 28. Ando, K., Hiroishi, K., Kaneko, T., Moriyama, T., Muto, Y., 14. Cui, J., Shin, T., Kawano, T., Sato, H., Kondo, E., Toura, I., Kayagaki, N., Yagita, H., Okumura, K. & Imawari, M. (1997) Kaneko, Y., Koseki, H., Kanno, M. & Taniguchi, M. (1997) J. Immunol. 158, 5283–5291. Science 278, 1623–1626. 29. Yu, Y. Y., George, T., Dorfman, J. R., Roland, J., Kumar, V. & 15. Taniguchi, M., Koseki, H., Tokuhisa, T., Masuda, K., Sato, H., Bennett, M. (1996) Immunity 4, 67–76. Downloaded by guest on September 30, 2021